Modeling Developmental Trajectories of White Matter Microstructure in the Autistic and Typical Brain
Objectives: Using diffusion tensor imaging (DTI) data from an ongoing longitudinal study of ASD, we investigated growth models of longitudinal white matter microstructure development and assessed which models best represent the developmental trajectories of 48 white matter tracts in individuals with ASD and TDC.
Methods: Male participants (N = 154; 99 ASD) between the ages of 3 and 52-years-old were scanned at the University of Utah up to 4 times across 9 years. The 48 white matter tracts defined in the JHU ICBM-DTI-81 template were aligned to a population-specific template created after image processing then median diffusion parameters were extracted for each tract. We fit the regional DTI data (i.e. FA, MD, RD, AD) to longitudinal linear and quadratic growth models and assessed which model best represented each tract’s development using the Bayesian Information Criterion parsimony metric. These best-fit models were compared between groups to examine whether white matter development in ASD and TDC follow different trajectories.
Results: The Bayesian Information Criterion revealed 80 instances across the 4-diffusion metrics (FA, MD, RD, AD) where the best-fit models differed between ASD and TDC. In particular, 18, 19, 19, and 24 of the 48 white matter tracts differed in their FA, MD, RD, and AD trajectories, respectively. In all but 3 of these cases, which included MD and RD in the fornix and AD in the left medial lemniscus, a linear growth model was found to better fit the TDC data while a quadratic model best fit the ASD population.
Conclusions: Quantitative analysis of white matter maturation may provide insight into processes that are altered during neurodevelopment in ASD. These preliminary findings suggest that the growth models used to represent the trajectory of white matter microstructure development differs between individuals with and without ASD. Notably, these differences appear to be regionally dependent as the trajectory differences were not observed in all investigated tracts. Further investigation into additional models, such as logistic or Gompertz growth models, is required to ascertain which model best characterizes the developmental trajectories of white matter. Future analyses will examine these additional models to assess whether they provide a more accurate representation of development and whether these models differ between ASD and TDC trajectories.